Meta is no longer treating its data centers as just background support. The company just launched Meta Compute, a new top-level organization designed to centralize everything related to AI hardware, energy, and supply chains. Mark Zuckerberg announced the move this week, making it clear that the ability to secure massive amounts of power and silicon is now a core part of Meta’s competitive strategy. The goal is to build out “tens of gigawatts” of capacity before the end of the decade, with plans to reach hundreds of gigawatts in the years following.
The scale of this project is difficult to overstate. A single gigawatt can power roughly 750,000 homes, so Meta is essentially planning to build an energy footprint comparable to a small country. To manage this, Meta Compute is bringing software, custom silicon design, and physical data center construction under one leadership team. This structural shift is meant to stop the company from hitting a “physical ceiling” where its AI ambitions are held back by a lack of electricity or land. It follows recent deals Meta signed with nuclear energy firms like TerraPower and Oklo to secure clean, reliable power for its newest clusters.
The initiative is being co-led by Santosh Janardhan, who handles the technical architecture and global data center operations, and Daniel Gross, who is focused on long-term capacity planning and supplier partnerships. They are also working with Dina Powell McCormick, a former Goldman Sachs executive and national security adviser, to coordinate with governments and sovereign entities on financing and building these massive infrastructure projects. By treating power and hardware as a single coordinated problem, Meta hopes to avoid the reactive, last-minute expansion that has characterized the AI arms race so far.
While Wall Street is watching the high price tag—Meta spent over $70 billion on AI infrastructure in 2025 alone—Zuckerberg is betting that “personal superintelligence” for billions of people will only be possible with this level of investment. You can expect to see more news about Meta’s custom chips and dedicated energy grids as these gigawatt-scale projects break ground over the next few years. For now, the focus is on securing the supply chain to make sure the hardware and the power are ready when the next generation of AI models needs them.
